(OLD) (ARCHIVED) Puppy Linux Discussion Forum

[emil]

Saluki (also Wary and Racy?)
R-2.14.1-i486.pet (18 MB), R_DEV-2.14.1-i486.pet (88 K), R_DOC-2.14.1-i486.pet (14 M), R_NLS-2.14.1-i486.pet (733 K), R-2.14.1-Saluki.sfs (32 MB)
Update 24. Jan 2012

Slacko RC (Puppy 5.3)
Download http://boxen.math.washington.edu/home/emil/MathSoftware/Puppy5-slacko/R-2.13.2-slacko.sfs

Lupu 511
2 Pets : R and R + Rcmdr
1 sfs : R-Rcmdr-2.12.1-lup511.sfs

Download http://boxen.math.washington.edu/home/emil/MathSoftware/Puppy5


Those were build in Slacko RC and Lupu Quickset Edition 511. sfs package also briefly tested in new Puppy 5.2.

The R-Rcmdr-2.12.1-lup511.sfs (145 mB) has built in gfortran 4.4.3 and TclTk8.5. It installs to /opt/R-2.12.1 (possible to include in Remaster).
The Rcmdr pet packages install to /mnt/home (no remaster possible, but usable from different frugal installs).

recommended/optional Packages:
devx.sfs (to install and compile additional packages)
tcltk8.5 (graphical download menu packages needs it)
Xorg-high.pet (from Quickpet for 3d plots)
gfortran

Just R: for those who just need commandline: start with R from terminal, installs normal
Rcmdr: R + Rcmdr, for those who like a gui. R - starts just R in terminal, Rcmdr starts R + Rcmdr, has also menu entries in Menu/Business/Calculate, installs to /mnt/home

The sfs has just an Menu entry for R, start Rcmdr with library(Rcmdr) from R prompt.

emil


--- Old Entry, ca Jan 2010, R 2.10.2 made in Puppy 431 ---
R seems to be standard package for statistical problems.
I compiled the latest version, download it:

R-2.10.1-i686.pet

Klick the pet then type R in a console (upper case).

I havent tested it much, so please report any problems.

I think it is possible to load additional packages from within R, but you will need the devx431.sfs development package to use them.

emil

[cullenjd]

Hi Emil,

Thanks for putting this package together. I'm new to linux in general and puppy linux specifically. When I tried to use the package I got the following error.

# R
/usr/local/lib/R/bin/exec/R: error while loading shared libraries: libgfortran.so.2: cannot open shared object file: No such file or directory

Any help you could provide will be appreciated.

John

[disciple]

There is a .pet at http://www.4shared.com/file/193573393/57ec0302/gfortran-422.html
Also needs iconv I think http://www.4shared.com/file/193574839/dc54d5ce/libiconv-250.html

[Puppyt]

Cheers again, emil.

I'm writing in relation to cullenjd's experience below, and also to my long-lamented request for an iron-clad pet or even SFS of R (see my early experiments at http://www.murga-linux.com/puppy/viewtopic.php?t=43221).
Do you think it worthwhile or possible to develop an R + Rcmdr (a GUI that moves R from a command-line environment to that of a spreadsheet one more familiar to beginners in statistics - think PSPP, StatView, SAS, Minitab: http://socserv.mcmaster.ca/jfox/Misc/Rcmdr/). Something with all the dependencies included (bet it wouldn't be as big as your GRETL pet!). I think to would be of immense help to people who are both new to Linux and to statistics, and who will naturally baulk at also learning the R scripting language - especially if they're busy with a university course etc.

Would anyone with expertise in compiling SFS's like to take this on?

[emil]

Hello,

Hum hum, I sweared to quit this math stuff. Anyway, I have Rcmdr working.

Rcmdr HowTo
1) dependencies
First of all there is a tarfile with with some dependencies (sorry, again on filehosting.org).
tcl-8.5.5-runtime.pet
tk-8.5.5-runtime.pet
LibXt-6.0.0.pet
Download (sorry - made a little mistake, no .tar extension with this file. Please rename it to Rcmdr_stuff.tar, then unpack it by klicking on it in the Rox Filemanager)
install those pets.

2) install R Maybe you have already done this, if not:
Second step, install the R pet, the libiconv pet and the gfortran pet.
(All from this thread or from Math Base thread)

3)install devx_431.sfs
Then you need the devx_431.sfs installed (from the official repository), because you will need the Gcc /g++ compilers when updating R.
Frugal install: copy it to your /mnt/home/ directory, config your Boot Manger from Menu/System and then rebot.

Full hard install: (Howto from official site)
Currently, installing the "devx" file in a Type 2 (full) hard drive installation is more complicated, and you can't undo it. It involves some steps....
1. download dev431.sfs from http://distro.ibiblio.org/pub/linux/distributions/puppylinux/puppy-4.3.1/
2. Click on the devx_xxx.sfs in a ROX-Filer window to mount it.
3. Open a terminal in the mounted directory.
6. # cp -a --remove-destination ./* /mnt/hda2/ <-- replace hda2 with your drive name (same as drive icons)
7. # sync
8. Close the terminal.
9. Click on the devx_xxx.sfs file to unmount it

Note, the '--remove-destination' option is essential. If you only use '-f' to force overwrite, it will follow (dereference) a symlink, that can cause unexpected overwrites. (You probably know all of this, but I want to keep it beginners friendly and, honestly, I have to look it up myself all of the time)

4) install OpenGL Dev package
Ok then it is importent to use Puppies package manager to install the
xorg_xorg_full_dri-7.3
xorg_xorg_full_dri_DEV-7.3
packages (R uses open GL for 3d plots)

5) Use R package manager
Ok, now you can start console and type R - this should boost you to the R environment.

Then invoke the R package managment via install.packages("Rcmdr", dependencies=TRUE) - original instructions here
Well, magic started, it should now compile lots of packages to update R, should run without problems.

6) start Rcmdr
at last, invoke Rcmdr by typing library (Rcmdr)

hell off stuff, but I hope it works, dont really have time to test it

@ Puppyt: I feel the same - puppy could be a scientific/educational rocket launcher on USB. Rcmdr at the moment is no easy pet, probably will never be. Best would be to compile something like a Mathpup II, but I suggest those packages should get testet now for some time and if they work reasonable well I might find the time to compile them all into an iso. I might ignore bug reports for the next weeks though. Issues in real life are calling. But feel free to post your experiences nonetheless.

Servus
emil

5.474 The number of fundamental operations that are necessary depends solely on our notation. - Tractatus
my place

[Puppyt]

emil - you are an absolute legend, thank you!

I really appreciate very much all the consideration and effort you have made to not only testing the R + Rcmdr package out, but also collecting the resources AND writing up such a detailed 'how to' with screen shot -
Marvelous stuff indeed, and it will be a treasured resource for others like myself treading the rocky paths of learning in the worlds of Linux and statistics.

Your points on the difficulties of constructing a pet (or SFS) of such a package is well made, and you underscore the need for a MathPup II. (Perhaps with all the mathematical goodies that you have been amassing on your other thread.) There are other R GUI's of note to beginners in stats and R - like Rkward (KDE) and JguiR (java) - but the extra dependencies that would be required might make such a mathematical pupplet rather large. For my own part I've meddled with the latest R deb package on gposil's Dpup477 (retro), but thank you again for the documentation and links. It's rekindled my enthusiasm for the cause and I'll have another testing exploration, using your instructions in the coming weeks.

Cheers!

[disciple]

The gtk gui "pmg" is probably worth checking out too.
http://wiener.math.csi.cuny.edu/pmg/2008/05/17/

[emil]

PMG - Poor Mans GUI - should go well with frugal installs.

It can be launched like this:
Like with Rcmdr you need the development/compile environment
installed (devx_xxx.sfs, gfortran ). Then I modified the install script, it is attached. Download it to a convenient location. Open a terminal there, untar and then launch R. Then type

source("pmg_install.R")


it should install and build packages, if it finishes launch pmg with

library ("pmg")

Servus,
emil
----
5.474 The number of fundamental operations that are necessary depends solely on our notation. - Tractatus
my place

[emil]

It is maybe also a good idea to post a subtle hint at the incredible


R-seek-machine (thanks @ carolus)

[disciple]

[emil]

hello disciple, thank you for that typo. It is library(pmg) of course.
Spelling is not my strong side (*sigh*). I'll correst it in the previous post.
I have no time to do in depth tests of those programs. Any reports on experiences would be welcome.
thank you
emil

[Puppyt]

Wow! I'd never heard of pmg - thanks very much for the heads-up, disciple, and thanks emil for the quick 'how to' on its implementation.

Looks like there are plenty of approaches under development to assist students learn their way into statistics via the R environment. I have to admit however that the comparatively underpowered, and obsolete StatView5.1 was my favourite learning tool - and many researchers in my field still swear by it, but not publicly (they're the ones who don't explicitly say what stats software they used in their journal publications). But as I mentioned in an earlier post, 'R' is the new black for increasing numbers of high-end research publications in the field of animal behaviour and comparative psychology, and it would be understandable to see similar trends in other fields.

I haven't compiled any of these yet to test - I imagine jrb's 431barebones could be a good base for USB-type rocket launcher, or even zenfunk's Pulp 0.1 or synth's TurboPup Xtreme 1.1 - when either of those offerings hit the dancefloor. Meantimes I imagine no headaches with compilations on any puppy4 derivative following your clear instructions.

Cheers!

[izreflexxx]

I got this error, when i tried to typed in source("pmg_install.R") :

> source("pmg_install.R")
Installing system libraries.
Loading Tcl/Tk interface ... Error in dyn.load(file, DLLpath = DLLpath, ...) :
unable to load shared library '/usr/local/lib/R/library/tcltk/libs/tcltk.so':
libtk8.5.so: cannot open shared object file: No such file or directory
Error: .onLoad failed in 'loadNamespace' for 'tcltk'

[disciple]

Try this http://www.filehosting.org/file/details/120921/pmg_and_deps.pet
~9MB
It includes these libraries:

bitops
cairoDevice
graph
gWidgets
gWidgetsRGtk2
pmg
proto
RBGL
RGtk2
SparseM
XML

I'm not actually sure that they are all really dependencies - the build script still wanted to install the R tcl/tk library even though it isn't needed.
N.B. I removed the RGtk2 help to make the package 6MB smaller.

[disciple]

BTW that error message was caused by the build script using the stupid "menu" function, which like other stupid R scripts wants to use the tcl/tk interface by default, even when it isn't available. You should be able to avoid it by adding a